For example, when I say "butter" with my North American English accent, I typically say the -tt- part with an voiced alveolar tap. It's represented by "ɾ" in IPA. This sounds close to a "d" sound, but faster. However, if you ask me to slow down and repeat "butter", I'll likely respond with a full on [voiceless alveolar plosive], represented by a "t" in IPA. (Look familiar?) Both [ɾ] and [t] are allophones of the /t/ morpheme in NA English. They're specific sounds that are perceived of as a "t".
Now, I am not an expert on Hindi at all, so I cannot make any specific claims myself regarding particular sounds, but if Wikipedia is reliable here, an example of similar allophony is [w] and [v].
* Side note, I'm condensing a bunch of things in here, so it leaves out some complexity, but it should hopefully express the point that no language will have all of its sounds represented perfectly in orthography.
Voice Alveolar Taps - https://en.wikipedia.org/wiki/Voiced_dental_and_alveolar_tap...
Voiceless Alveolar Plosives - https://en.wikipedia.org/wiki/Voiceless_dental_and_alveolar_...
w/v allophony in Hindi - https://en.wikipedia.org/wiki/Hindustani_phonology#Allophony...
Yes schwa deletion is a counter example, although it's a solid rule so you could easily say 'well that's just one thing to learn on top of the script, that it's written but not pronounced in that placement'.
A better case for 'no not 100%' IMO is the behaviour of the 'h' sound, which converts 'aha' ('a's around it) it to sound something more like 'ehe'. बहन for example means sister, if you transliterated according to IAST it would be bahan, but most would write behen, and that's how it's pronounced (or something between e and ai anyway).
https://old.reddit.com/r/IndiaSpeaks/comments/hrl6lp/comment...
I thought to check because if you look into Spanish for example, it is often thought to be phonetic but it really isn't 100% phonetic either.